Fingerprinting arguments, first introduced by Bun, Ullman, and Vadhan (STOC 2014), are the most widely used method for establishing lower bounds on the sample complexity or error of approximately differentially private (DP) algorithms. Still, there are many problems in differential privacy for which we don’t know suitable lower bounds, and even for problems that we do, the lower bounds are not smooth, and usually become vacuous when the error is larger than some threshold. In this work, we present a new framework and tools to generate smooth lower bounds on the sample complexity of differentially private algorithms satisfying very weak accuracy. We illustrate the applicability of our method by providing new lower bounds in various settings: 1. A tight lower bound for DP averaging in the low-accuracy regime, which in particular implies a lower bound for the private 1-cluster problem introduced by Nissim, Stemmer, and Vadhan (PODS 2016). 2. A lower bound on the additive error of DP algorithms for approximate k-means clustering and general (k,z)-clustering, as a function of the multiplicative error, which is tight for a constant multiplication error. 3. A lower bound for estimating the top singular vector of a matrix under DP in low-accuracy regimes, which is a special case of DP subspace estimation studied by Singhal and Steinke (NeurIPS 2021). Our main technique is to apply a padding-and-permuting transformation to a fingerprinting code. However, rather than proving our results using a black-box access to an existing fingerprinting code (e.g., Tardos’ code), we develop a new fingerprinting lemma that is stronger than those of Dwork et al. (FOCS 2015) and Bun et al. (SODA 2017), and prove our lower bounds directly from the lemma. Our lemma, in particular, gives a simpler fingerprinting code construction with optimal rate (up to polylogarithmic factors) that is of independent interest.

There is a huge gap between the upper and lower bounds on the share size of secret-sharing schemes for$n$-party access structures; consistent with our current knowledge the optimal share size can be anywhere between polynomial and exponential in$n$. For linear secret-sharing schemes, the share size for almost all$n$-party access structures is exponential in$n$. We would like to study larger classes of secret-sharing schemes with two goals: 1) prove lower bounds for larger classes of secret-sharing schemes; and 2) construct efficient secret-sharing schemes. Given this motivation, Paskin-Cherniavsky and Radune (ITC’20) introduced a new class of secret-sharing schemes in which the shares are generated by applying degree-$d$polynomials to the secret and some random field elements. We define and study two additional classes of polynomial secret-sharing schemes: 1) schemes in which the reconstruction of the secret is done using polynomials; and 2) schemes in which both sharing and reconstruction are done by polynomials. Our main result is a construction of secret-sharing schemes and conditional disclosure of secrets protocols with quadratic sharing and reconstruction that are more efficient than linear secret-sharing schemes. To complement our results, we prove lower bounds on the share size for schemes with polynomial reconstruction. Finally, we give an evidence that schemes with polynomial sharing are probably stronger than schemes with polynomial reconstruction.

A secret-sharing scheme realizes the forbidden graph access structure determined by a graph$G=(V,E)$if the parties are the vertices of the graph and the subsets that can reconstruct the secret are the pairs of vertices in$E$(i.e., the edges) and the subsets of at least three vertices. Secret-sharing schemes for forbidden graph access structures defined by bipartite graphs are equivalent to conditional disclosure of secrets (CDS) protocols. We study the complexity of realizing a forbidden graph access structure by linear secret-sharing schemes, which are schemes in which the secret can be reconstructed from the shares by a linear mapping. We provide efficient constructions and lower bounds on the share size of linear secret-sharing schemes for sparse and very dense graphs, closing the gap between upper and lower bounds. Given a sparse (resp. very dense) graph with$n$vertices and at most$n^{1+\beta }$edges (resp. at least$\binom {n}{2} - n^{1+\beta }$edges), for some$0 \leq \beta < 1$, we construct a linear secret-sharing scheme realizing its forbidden graph access structure with total share size$\tilde {O} (n^{1+\beta /2})$. Furthermore, we construct linear secret-sharing schemes realizing these access structures in which the size of each share is$\tilde {O} (n^{1/4+\beta /4})$. We also provide constructions achieving different trade-offs between the size of each share and the total share size. We prove that almost all forbidden graph access structures require linear secret-sharing schemes with total share size$\Omega (n^{3/2})$; this shows that the construction of Gay, Kerenidis, and Wee [CRYPTO 2015] is optimal. Furthermore, we show that for every$0 \leq \beta < 1$there exist a graph with at most$n^{1+\beta }$edges and a graph with at least$\binom {n}{2}-n^{1+\beta }$edges such that the total share size in any linear secret-sharing scheme realizing the associated forbidden graph access structures is$\Omega (n^{1+\beta /2})$. Finally, we show that for every$0 \leq \beta < 1$there exist a graph with at most$n^{1+\beta }$edges and a graph with at least$\binom {n}{2}-n^{1+\beta }$edges such that the size of the share of at least one party in any linear secret-sharing scheme realizing these forbidden graph access structures is$\Omega (n^{1/4+\beta /4})$. This shows that our constructions are optimal (up to poly-logarithmic factors).

A secret-sharing scheme allows to distribute a secret s among n parties such that only some predefined “authorized” sets of parties can reconstruct the secret, and all other “unauthorized” sets learn nothing about s. For over 30 years, it was known that any (monotone) collection of authorized sets can be realized by a secret-sharing scheme whose shares are of size 2 n−o(n) and until recently no better scheme was known. In a recent breakthrough, Liu and Vaikuntanathan (STOC 2018) have reduced the share size to 20.994n+o(n), which was later improved to 20.892n+o(n) by Applebaum et al. (EUROCRYPT 2019).